Dynamic control of deep-level mine cooling systems using artificial intelligence

Abstract

Deep-level mines face high electricity costs, posing a challenge as it significantly impacts production costs, consequently decreasing overall profitability. Complex energy-intensive cooling systems, including ventilation and refrigeration, contribute to approximately 28% of the mine’s total energy consumption. Hence, implementing energy savings initiatives for cooling systems in deep-level mining can mitigate production costs in the industry. The existing approaches to reducing electricity consumption on mine cooling systems rely on static control strategies and fail to adapt to the dynamic conditions prevalent in deep-level mines. The mining industry has rarely embraced time series forecasting models using artificial intelligence (AI) to develop control strategies for cooling systems. This study introduces AI-based temperature forecasting to address these limitations to enable the development of a dynamic control strategy for cooling systems in deep-level mines. By leveraging AI, the proposed approach aims to overcome the shortcomings of the existing methods and improve energy efficiency in the mining industry. The proposed dynamic control strategy focused on the surface bulk air cooling (BAC) plants and fans of a deep-level gold mine, considering system parameters closely related to the underground conditions and the operating characteristics of the system components. Long short-term memory recurrent neural network (LSTM-RNN) models were employed with five different forecasting horizons to forecast the surface and underground air temperatures. The accuracy of the temperature forecasting model was validated against real data and previous literature. The best-performing model (30-minute forecast) yielded a coefficient of variation for the root mean square error (CV(RMSE)) of 1.66% and 1.55%, a root mean square error (RMSE) of 0.40°C and 0.37°C and a mean absolute percentage error (MAPE) of 2.29% and 1.01% for the surface air wet bulb temperatures and underground wet bulb temperatures. These findings confirm that the model accurately forecasts surface and underground air temperatures. Subsequently, the model was used to identify a dynamic control strategy for the surface BAC system components, further enhancing the system's energy efficiency. A possible electricity cost savings of ZAR 1.5 million per annum were realised based on the proposed strategy. The models used to determine the control strategies only required significant maintenance when an unforeseen event occurred, such as adding or removing a fridge plant. Implementing such models aids the mine in achieving electricity cost savings without any additional changes to the current infrastructure. However, implementing an automated system will supplement the existing savings by making factors such as human error inconsequential.